Abstract: Different forests show various spectral characteristics because of the differences in canopy structure and functional traits.Analysis of spectral characteristics during different forests can provide a basis for high precision forest classification and canopy parameter inversion.Therefore,a clear understandingof the hyperspectral characteristics of typical forest types in China isuseful toimprove the accuracy of forest classification and forest physical and chemical parametersinversion.Four typical forest sites (Changbai Mountain,Shennongjia,Qianyanzhou and Dinghushan) of Chinese Ecosystem Research Network (CERN) have been studied.The reflectance spectrum curves of four kinds of forests have been extracted in pixel scales from EO-1 Hyperion images.First-order differential has been employed to reduce noise from spectral data and some indexes have been calculated to quantitatively analyze its spectral characteristics.The results show that firstly,the reflectance of broadleaved mixed forest at Shennongjia is significantly higher than that of the other three forest types in the whole spectral range.Coniferous mixed forest at Qianyanzhou shows a low reflectance.Changbai Mountain and Dinghushan forests own medium reflectivity.Secondly there are significant differences in the red edge slope during the four forest types.The red edge slope of Shennongjia forest is the highest,suggesting that its canopy chlorophyll content is the highest and the best health.Compared with the other three type forests,the red edge position of Qianyanzhou forest had red-shift and its red edge is located at 720nm.In addition,according to remote sensing indexes,Normalized Difference Nitrogen Index (NDNI) of broadleaved mixed forest at Shennongjia is larger than that of coniferous mixed forest at Qianyanzhou,which reflects the difference of leaf nitrogen concentration between them.In addition,due to high forest coverage,Enhanced Vegetation Index (EVI) is likely to indicate accurate changes in forest greenness and coverage than Normalized Difference Vegetation Index (NDVI).This study is very important for remote sensing inversion of forest parameters and high precision forest classification in China.